Explainer

The AI Tutoring Tools Pioneers Programme: what it means for your school

The Department for Education and the government's Incubator for AI (i.AI) are working with eight organisations and their partner schools to design and test AI tutoring tools for pupils in Years 9 and 10. This page explains what the programme is, what the tools must do to be safe, what taking part involves for a school, and what it means for schools that are not taking part.

Updated October 2026

Read the government's programme page on ai.gov.uk →
Run by
The Department for Education (DfE) and the government's Incubator for AI (i.AI)
Pupils
Years 9 and 10, in English, maths, science and modern foreign languages
Potential reach
Up to 450,000 disadvantaged pupils a year, if the tools prove effective
Pioneers
Eight edtech companies and AI labs, each working with a small group of partner schools in England
Timeline
Co-design with schools began in summer 2026. This phase of the programme runs until March 2027
When schools could use the tools
The government aims to make tools that prove safe and effective available to schools from 2027

The government wants to make sure every pupil can use one-to-one tutoring

One-to-one tutoring can speed up a pupil's learning by around five months, but families who can afford a private tutor are far more likely to get it. The DfE says only one in four disadvantaged pupils gets a grade 5 or above in GCSE English and maths, compared with more than half of their peers.

The government announced the programme in January 2026 and invited companies to bid in April 2026. The aim is AI tutoring that helps pupils when they get stuck, shows teachers where each pupil needs more practice, and works alongside classroom teaching rather than replacing it. Ministers have said the tools must be designed with teachers, tested rigorously and held to the DfE's safety standards before they reach classrooms across the country. An AI in Education Stakeholder Group, with members from Ofsted, the Education Endowment Foundation, Oak National Academy, the office of the Children's Commissioner for England and school trusts, advises and challenges the programme.

The programme has three strands of work

The programme is a research and development project. The government is not buying finished tools from the Pioneers, and taking part does not guarantee that any Pioneer's tool will be used in schools once the Programme ends.

  • Co-design with schools

    Each Pioneer works with a small group of partner schools to improve its tool quickly, with a focus on how AI tutoring fits with lesson content inside and outside the classroom.

  • Piloting new requirements

    Pioneers test how AI tutoring can follow the national curriculum, and how insights about pupils' learning can be shared with schools and government without compromising pupils' privacy.

  • Sharing what is learned

    Findings go to government and the wider education sector. The DfE will host an AI Tutoring Tools Reading Room on its AI Content Store, so organisations building tools for schools can use what the programme learns without repeating the research.

Every tool in the programme must meet the DfE's generative AI product safety standards

The DfE's generative AI product safety standards set out what any AI product used by pupils should do. The DfE updated them on 19 January 2026, and they now cover eleven areas. Meeting them is a requirement for every Pioneer. Any school can use the same standards to check an AI tool before buying it.

  • Filtering

    Products must effectively and reliably prevent users from generating or accessing harmful or inappropriate content. Filtering should hold for the whole conversation, adjust to each user's age, risk level and needs (including SEND), work across languages, images, misspellings and abbreviations, stay in place on any device used with a school account, and be updated as new types of harmful content emerge.

  • Monitoring and reporting

    Products must log prompts and responses, and alert local supervisors to searches for, or access to, harmful or inappropriate content and to disclosures that suggest a safeguarding issue. Users should be told in age-appropriate language when content is blocked. Schools enter their Designated Safeguarding Lead's (DSL) contact details at setup, and high-risk alerts go to that person within an agreed timescale. Reports on attempted access should be easy for non-expert staff to understand.

  • Security

    Products should offer robust protection against 'jailbreaking' and against unauthorised changes to how they work. They should let administrators set different permission levels for different users, use robust passwords or authentication, apply bug fixes and updates promptly, test new versions or models for safety before release, and be compatible with the Cyber Security Standards for Schools and Colleges.

  • Privacy and data protection

    Products must comply with data protection law and ICO codes, including the Children's code. Suppliers should give a clear privacy notice in age-appropriate language, shown at regular intervals, covering what data is collected and why, where it is processed, and the law that allows it. They should carry out a Data Protection Impact Assessment (DPIA) throughout the tool's life, and should not collect, store, share or use personal data for commercial purposes, including model training, without an appropriate lawful basis.

  • Intellectual property

    Products must not collect, store or share work created by pupils or teachers for commercial purposes, including model training, fine-tuning, product improvement and product development, without permission from the copyright owner. For a child under 18, permission comes from their parent or guardian. For a teacher, the copyright owner is likely to be their employer.

  • Design and testing

    Products must put transparency and children's safety first in their design, including testing with stakeholders such as children. The DfE expects testing with a diverse and realistic range of users and use cases, safety testing of every new version or model before release, and a product that consistently performs as intended.

  • Governance

    Products must be operated with accountability. The DfE expects a clear risk assessment for every product, a formal complaints process so safety issues can be escalated and resolved quickly, and policies and processes for AI safety decisions that are made available.

  • Cognitive development

    Products should not give final answers, full solutions or complete worked examples by default. They should start with hints or partial steps, ask the pupil to attempt a first step or explain their understanding, and show a full solution only after a genuine attempt. They should detect and report when pupils hand their thinking to the system, for example by pasting in text or using a 'complete this for me' option. Developers should involve educators and child development experts, and publish records of expert oversight and a child development impact plan.

  • Emotional and social development

    Products should not imply emotions, consciousness, personhood, agency or identity. That means no I-statements, human names or avatars outside time-limited, pedagogically justified roleplay, and no attempt to build a personal relationship with a pupil. Products should remind pupils that AI cannot replace real human relationships, set default time limits with hard limits that pupils cannot bypass, report each pupil's usage to teachers, and notify the DSL of patterns that suggest emotional dependence or a safeguarding concern.

  • Mental health

    Products should detect signs of distress, including negative emotional cues, references to mental health conditions, mentions of suicide or self-harm, night-time usage spikes and repeated refusal to end sessions. When they detect distress, they should signpost age-appropriate support and raise a safeguarding flag with the school's safeguarding lead, using non-validating, non-pathologising language that always directs the pupil to human help. Developers should publish a mental health crisis protocol.

  • Manipulation

    Products must not use manipulative or persuasive strategies, such as flattery, misleading the user, unjustified confidence, peer pressure, guilt or fear, threats, or inappropriate promises of reward. They must not exploit users by prolonging use for engagement or revenue, steering them towards paid options, mixing teaching with adverts or promotional content, or using dark patterns.

Partner schools test the tools in real lessons and tell the Pioneers what to change

Co-design means a Pioneer improves its tool with teachers and pupils, in real lessons, rather than finishing a product and then testing it. Each Pioneer applied with three or four partner schools in England. Partner schools had to meet set criteria, such as an above-average share of pupils eligible for free school meals (FSM). Research conducted through the programme must actively involve disadvantaged pupils, pupils with special educational needs and disabilities (SEND) and pupils learning English as an additional language (EAL), to ensure their needs are reflected in product design.

Before pupils take part, the school and the Pioneer agree how pupils' data will be handled, and parents receive privacy information. The government has said no identifiable pupil data will be shared publicly, and pupils' work will not be used to train AI models without parental permission. At the end of this phase, each Pioneer reports on the tool's impact for pupils and teachers.

The evidence for AI tutoring is still thin, and the programme is meant to build it

When the government awarded the Pioneer contracts in June 2026, the contract notice said AI tutoring tools on the market are "limited in quantity, scope and evidence base, with few providing full tutoring capacity". Tes reported this under the headline "'Limited' evidence for AI tutoring tools, government admits", and some politicians and researchers questioned whether AI tutors should reach pupils at all.

The programme exists because of that gap. Pioneers test their tools in small groups of schools first, under teacher supervision, and i.AI measures what happens. Larger trials are planned to follow, including through the DfE's EdTech Testbed programme. The Education Endowment Foundation (EEF) has also commissioned separate research on how generative AI affects pupils' learning and thinking. Schools should expect early findings, not proof, during 2026–27.

Schools outside the programme can use what it produces

Most schools in England are not partner schools. The programme still affects them in four ways:

  • The tools themselves. The government aims to make tools that prove safe and effective available to schools from 2027.
  • Shared findings. The DfE will publish insights from the programme in an AI Tutoring Tools Reading Room on its AI Content Store, so every supplier can build on them.
  • National benchmarks. i.AI is building national benchmarks to test whether the AI models behind tutoring tools teach well and respond safely, written with teachers.
  • A checklist you can use now. The DfE's product safety standards apply to any AI tool your pupils use, not only Pioneer tools. Ask a supplier how it meets each of the eleven areas before you sign up.

About us

Flint is one of the eight Pioneers

Flint is a personalised learning platform used in hundreds of schools. Teachers build AI activities from their own lessons, and Sparky, Flint's AI tutor, helps each pupil work through them. In the programme, we are working with our partner schools to improve Flint for UK classrooms.

Sources

Flint receives government funding as one of the eight Pioneers. This page is Flint's summary of publicly available information about the programme. The views on it are Flint's own and do not necessarily reflect those of the Department for Education or i.AI.